Adverse environmental impacts from deforestation are a growing area of concern in climate change discussions. The Chinese government has implemented a series of policies, such as the Grain for Green Project, in an attempt to mitigate the impacts. This study takes a regional perspective to report land use/land cover changes over the Loess Plateau region from 2001 to 2009. MODIS data were used in analyzing both the conversions among and the resulting changes in different land types. Government statistical census data and observed climate data were also incorporated in the analysis. A general consistency is shown in both remotely sensed and census data. With the implementation of various projects, including the Grain for Green Project, the total areas covered by grassland, cropland, and forests have increased by 19.2 % (6.05 × 104 km2), 33.7 % (5.80 × 104 km2), and 19.6 % (3.08 × 104 km2), respectively, during the 9-year period. While climatic conditions, particularly annual precipitation totals, usually dominate the distribution of vegetation, it is found that socioeconomic polices and human activities contribute to the increase in overall greenness and to vegetation growth (e.g., LAI increased by 16.8 % (0.10) overall). However, the feedback of land use/land cover to regional climate is complicated and cannot be easily distinguished from natural climate variations based on short-term observational data alone. To better isolate the effects, further analysis and modeling studies are suggested.
To address various types of natural disasters and environmental crises, China has implemented large-scale reforestation/afforestation policies to contribute to a sustainable environment (e.g., Liu et al. 2008). The most well known government-funded program is the Grain for Green Project (GFGP), which focuses on the return of terrace-farming croplands to forests/grasslands with compensation from the state (e.g., Tao et al. 2004). This policy has resulted in large-scale redistribution of land-use rights and has been ongoing for more than a decade throughout China (Hvistendahl 2012). Some of the immediate benefits include increased vegetation coverage, soil erosion control, and reduced spread of wind-blown dust. However, additional longer-term and larger-scale ecological, environmental, and climatic impacts may also be occurring (Pielke 2005; Bonan 2008; Liu et al. 2008).
The Chinese government has promoted a policy encouraging afforestation since the 1980s. This developed into the adoption of a more proactive step through the enforcement of a pilot GFGP on the Loess Plateau in 1999, including the provinces of Sichuan, Shaanxi, and Gansu; a nationwide GFGP was enacted in 2003. China invested about Chinese currency RMB 190 billion (equivalent of USD 23.2 billion) over a 10-year period (1999–2008) to subsidize farmers in regions with poor agricultural production or severe land erosion problems to enforce implementation of the GFGP, with an additional RMB 10 billion (USD 1.4 billion) pledged through 2015 (National Bureau of Statistics of China 2009b). Since the initial implementation of the GFGP in 1999, many studies have focused on its impacts at scattered local and small-regional scales. For example, Cao et al. (2007) reported that, in northern Shaanxi province, afforestation, along with the removal of livestock from over-grazed land, resulted in a large increase in the spatial coverage of trees and grasslands. Most of the existing studies take the perspective of overall changes in greenness or vegetation coverage over the region (e.g., Xin et al. 2008; Li et al. 2013). However, due to the nature of the GFGP as described above along with other human-induced afforestation efforts, the LULC change (LULCC hereafter) may also take the form of conversion among vegetation types. This study focuses on investigating at plateau-scale of these conversions and how they relate to regional climate.
LULC data was retrieved from the Moderate Resolution Imaging Spectroradiometer (MODIS) (Strahler et al. 1999) on board the NASA Terra and Aqua satellites. MODIS LULC data include 17 classified land-cover types that were readily available for this study. This paper focuses on the changes to six major land-cover types, including bare ground, shrubland, cropland, grassland, woody savannah, and forest. These land types cover 98.5 % of the Loess Plateau study area. Such LULCCs may have been directly impacted by government policy over the last decade, though there has heretofore been a lack of study of such changes in and conversions among the different land types, particularly at larger spatial scales.
Analysis of modern remotely sensed LULC data to document changes in and conversions among land-cover types that are related to the GFGP policy in the Loess Plateau region over the last decade, and
Validation of satellite-derived LULC data with census data to investigate the quality of both data sources and their consistency in revealing the source of the changes.
MODIS data have been used to monitor terrestrial ecosystem processes due to their advanced radiometric and geometric properties (Strahler et al. 1999; Zhang et al. 2003, 2008). The MODIS land-cover data-product MCD12Q1 used in this study have a horizontal resolution of 500 m, and were available annually from 2001 to 2009.
In addition, MODIS-based leaf area index (LAI) data was used to quantify the coverage and changes of vegetation growth (NASA Land Processes Data Active Archive Center 2010; Yuan et al. 2011). LAI is defined as the total one-sided green leaf area per unit of ground area in broadleaf canopies, and as one-half total needle leaf surface area per unit of ground area in coniferous canopies. This dimensionless index is an indicator of the amount of available leaf material, and therefore reflects the canopy density and growth of vegetation (e.g., Kang et al. 2003; Hanes and Schwartz 2011). The LAI data have 1-km horizontal and 8-day temporal resolutions.
The other widely used index for vegetation greenness is the Normalized Difference Vegetation Index (NDVI), derived from the Advanced Very High Resolution Radiometer (AVHRR) on board the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites (Goward et al. 1991; Myneni et al. 1997; Hargrove et al. 2009). NDVI is an indicator of the vegetative greenness of an area and, therefore, changes in NDVI can reveal changes in vegetation. NDVI data were available from 1981 to 2006 at 8-km horizontal and 15-day temporal resolutions.
Observed precipitation and surface air temperature data from the China Meteorological Administration were used for regional climate analysis. Another important data source was the census data published annually by the National Bureau of Statistics of China (e.g., 2011, http://www.stats.gov.cn/tjsj/ndsj/2011/indexeh.htm) and developed through direct ground-based human survey. This data were compiled by multiple governmental agencies from county, province, and up to national level based on divisions of administrative areas. For LULC, there are categories of total coverage areas of forests, grassland, cropland, urban development, as well as resulting total area of forests from specific projects and agricultural productivities. The LULCC for different land-cover types are derived from the annual data.
3 Regional climate over the Loess Plateau
In addition, Fig. 3c and d show the time series of the annual mean precipitation and surface air temperature anomalies, respectively. Precipitation (Fig. 3c) exhibits periods of strong variability in the 1960s, mid- to late 1980s, and late 1990s through 2003. Other periods were close to normal. Overall, more dry years than wet years occurred over this time period. It also appears that a recovery from a late-1990s dry spell has occurred over the last decade. Long-term variability of precipitation anomalies, after the shorter-scale variabilities were filtered with a 7-year running mean, shows a periodic fluctuation with a period of about 15–20 years. The surface air temperature anomaly (Fig. 3d) shows a clear warming trend. However, the trend declined somewhat from the late 1990s to 2010. Although these results show consistency with the findings of Ma and Fu (2006) in that a long-term drying trend started in late 1990s, the extended study period suggests a rebounding feature during the last few years.
4 Land use/land cover change (LULCC) over the Loess Plateau
4.1 MODIS-observed changes
The interannual changes of the land covers show clear relationships with climatic conditions. For example, in addition to the warmer temperatures, precipitation gradually increased since 2000 and reached a peak in 2003 (Fig. 3c and d). The grassland, cropland, and forest show increases in areal coverage since 2000 and lasted about 2 years after 2003 (Fig. 5). A negative precipitation anomaly occurred in 2006. Correspondingly, the grassland, cropland, and forest areas decreased in 2006 and lasted into 2007. After the positive precipitation anomaly in 2007, a steady increase in these three land types took place since 2008. On the contrary, the coverage areas of barren land, shrubland, and woody savannah showed opposite changes.
A comparison between 2005 and 2000 NDVI is also shown in Fig. 6a and b. Based on the definition of LAI and NDVI, their relationship changes with the leaf phenological periods (Wang et al. 2005). During leaf emergence and leaf production (greening) periods, there is a strong linear relationship between NDVI and LAI. During vegetative growth period, NDVI reaches maximum when LAI is above certain level. During leaf senescence period in autumn, NDVI decreases quicker than LAI. Figure 6a and b show greater LAI increase in the summer of 2005 and 2009 in comparison to 2000 than NDVI mainly due to that the NDVI became saturated (reaching its maximum) during the vegetative growth period.
4.2 Human activity-related LULCCs
In the Loess Plateau region, the primary human-related activities that affect LULCCs include farming, grazing, deforestation, and afforestation/reforestation. With the advent of the GFGP, some of the abandoned croplands were fenced in order to prevent over-grazing (Wang et al. 2010). Another reason for the abandonment of croplands in the last decade relates to changes in socioeconomic conditions because most of the agricultural activities in poorly developed rural areas rely on manpower; however, many young people migrated to big cities as reported by Wei (2008). Human activity-related LULCCs can therefore be inferred from an analysis of changes in cropland and total forest coverage, as well as the conversions from/to other LULC types.
Figure 9 shows the areas of afforestation that resulted from five key projects in China and reported in the census data. These projects include: Preservation of Natural Forests, GFGP, Harnessing of Source of Sand and Dust in the cities of Beijing and Tianjin, Protection forests in North China and the Yangtze River Basin, and Fast-growing Timber Forests. Among the five projects, GFGP has contributed the most to newly forested areas. At the end of 2010, the accumulated total area of afforestation in China was about 6.4 × 105 km2 (64 million hectares), of which 37 % resulted from the GFGP (Fig. 9; National Bureau of Statistics of China 2011). Interestingly, the annual total afforested areas, as well as the accumulations, resemble the results from MODIS-observed changes from other types to forest, as shown in Fig. 8a. The correlation coefficient between the annual other-to-forest conversion in Fig. 8a (red bars) and the annual total afforestation in Fig. 9 (black bars) is R = 0.91, with a 99.0 % confidence level in Student’s t-test. It needs to be noted that the difference in magnitude between Figs. 8a and 9 is due to the fact that Fig. 8a represents results for the Loess Plateau study area alone, while Fig. 9 displays results for the entire country. In fact, the annual accumulations of the afforestation projects (curves in Fig. 9) exhibit even stronger correlations to the MODIS-observed accumulation of other-to-forest conversions (red curve in Fig. 8a). The correlation coefficient is R = 0.93 (99.9 % confidence level) for the annual accumulation of GFGP, and R = 0.99 (99.9 % confidence level) for the annual total accumulation of all five afforestation projects. These results suggest that the MODIS-observed LULCCs are reliable. In addition, the consistency between MODIS data and government survey data implies that both data sources can support and supplement one another in regional climate studies. Although, as Bradley and Mustard (2004) pointed out, some variability in land productivity may be inherent to a dominant ecosystem, the high correlation of forest area changes shown in Figs. 8a and 9 implies that these changes are due to the active LULCC resulting from government policy and human activities.
5 Summary and discussions
A warming trend occurred throughout the entire study area over the past 50 years (1961–2010), with the amplitude of warming increasing from south to north and ranging from 0.0 to 1.5 °C/10-year, and decreasing in the late 2000s. The annual precipitation showed a wetting trend in the arid west and a drying trend in the central to the southern parts of the study area. The maximum amplitudes were 43.1 mm/10-year for wetting and −39.5 mm/10-year for drying. In the past decade, the late 2000s were relatively drier than earlier in the decade.
While vegetation types over the Loess Plateau generally depend on the distribution of annual precipitation, with temperature representing a secondary factor, LULC changes were notably affected by socioeconomic policies and human activities, including farming and/or implementation of the GFGP. The increase in vegetation coverage as a result of the afforestation effort was significant, and was supported by both census data and remote-sensing data. For example, the accumulated change from cropland to other land-cover types during 2001–09 was 14.08 × 104 km2, representing 12.5 % of the entire study area, and the accumulated change from other land-cover types to forest was 3.54 × 104 km2, representing 2.9 % of the entire study area. The net increases in spatial coverage of grassland, cropland, and forest were 6.05 × 104 km2 (19.2 %), 5.80 × 104 km2 (33.7 %), and 3.08 × 104 km2 (19.6 %), respectively. A resulting net increase in LAI was 0.10 (16.8 %).
The interaction between regional climate and LULC is complicated, and yet is an important scientific subject that needs further attention (Pielke 2005; Mahmood et al. 2010). The impact of LULCC on climate is a long-debated question, as reviewed by Ellison et al. (2012). From this study we see the dominant role of climate on vegetation distribution and changes; however significant changes in LULC have occurred over the past decade due to human activities. Nevertheless, quantifying the feedback from changed LULC to the climate requires long-term data analysis. Furthermore, distinguishing the impacts of human activities from natural climate variations can be further investigated through the use of numerical modeling, which will help us to more fully understand how afforestation efforts contribute to regional and global climate. A preliminary study of the interactions between climate and LULC is presented in the accompanying work of Fan et al. (2014).
This research was supported by the National Basic Research Program of China (2012CB956201), the National Natural Science Foundation of China (41275085), the Knowledge Innovation Program of the Chinese Academy of Sciences (KZCX2-EW-202), and the Special Fund for Meteorological Scientific Research in Public Interest (GYHY201106028). Jeremy Krieger at the University of Alaska Fairbanks helped editing and proofreading the manuscript.
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.